Posts

Showing posts from May, 2025

TROY 180525 wins Elite Engines Tournament, 31.05.2025

Image
This is another big tournament (1360 games played) of elite chess engines. TROY 180525 won and SF-PRO202.05.2025 and Yuliana 7.0 engines split 2-3. More results soon.   CEDR  646.387 games download  (01.05.2025) 💾  Tournament download with 1360 games 💻Lenovo Z710 i7-4710MG 8GB RAM 🖬 GUI-Arena 👍 @chessenginesdiary Country - Poland, City -  Malbork   🕓 Time 3'+3"  Tech table: Engine KN/mov NPS dep/mov time/mov mov/gam time/gam fails Artemis 15TR 9777 1811756 46.5 5.4 66.9 361.2 Caissa 1.22 23101 4203307 41.0 5.5 65.9 362.0 Cool Iris v19 5561 1030556 43.6 5.4 66.0 356.1 1 DarkSeid v5.9 9489 1754825 46.8 5.4 65.0 351.4 DarkSeid v6.0 9715 1798688 47.6 5.4 66.2 357.5 Halogen 12.23.0 JA 22058 4440900 26.9 5.0 80.9 401.9 Killfish PB 200525 11136 2032370 49.0 5.5 63.1 345.8 Obsidian 16.0 14948 2734747 41.9 5.5 64.9 354.6 Obsidian dev-16.01 15535 2836644 41.9 5.5 64.0 350.3 Patricia 4.0 Dev.4cb JA 24106 4861127 33.8 5.0 81...

Update MEGA Dysk for donors - 31.05.2025 ( 11 files added)

Image
We've just launched a new MEGA drive for our sponsors! We are incredibly grateful for your crucial support. Your contributions are essential for the development of 'Chess Engines Diary' and enable us to keep delivering new content to you. Look: New Bonus Rules from the Chess Diary MEGA Dysk - update 31.05 .2025   (11 files added) Engines for Windows: (9 files):  Aku 2025.30.5 JA, BlackFish 270525, Cinder 0.2.0, DarkZoryn, Obdisian 16.02, Obdisian 16.03, Omega 6, Omega 6-M, Tarnished 2.1 JA. MacOS (1 files):  Cinder 0.2.0. --- Engines for Android (1 files):  Stockfish 250525. ---- Book and NNUE (0 files):--- --- Games (0 file):--- GUI (0 files): --- --- How to become a sponsor: -  here

Obsidian 16 wins Arena Strong Engines Test, 26-30.05.2025

Image
Another tournament held on an older, 10-year-old computer. New versions of powerful chess engines were playing. Obsidian 16 won, Quanticade Orion 2.0 in second place and Starzix 6.1 JA engine in third place. More results soon on our website. CEDR  646.387 games download  (01.05.2025) 💾  Tournament download with 450 games 💻Lenovo Z710 i7-4710MG 8GB RAM 🖬 GUI-Arena 👍 @chessenginesdiary Country - Poland, City -  Malbork   🕓 Time 3'+3"  Rank Engine CEDR Score % S-B 01 Obsidian 16 3737 65.0/90 72.2  2605,25  02 Quanticade Orion 2.0 3563 58.0/90 64.4  2276,75  03 Starzix 6.1 JA 3631 56.0/90 62.2  2260,50  04 Koivisto 9.2 JA 3615 55.0/90 61.1  2172,50  05 Uralochka 3.42 dev9 JA 3646 54.0/90 60.0  2130,75  06 Halogen 12.23.0 JA 3543 42.5/90 47.2  1649,75  07 Zangdar 4.01 2965 41.5/90 46.1  1529,75  08 Turbulence 0.0.7 2649 39.5/90 43.8  1475,25  09 Bagatur...

Obsidian 16.03 - new version strong chess engine

Image
Obsidian , a n UCI chess engine written in c++ Rating Chess Engines Diary  CEDR=3732 v.16: This new version includes a few search patches and a new NNUE architecture, finally with layers. I dropped support for pre-AVX2, but I might add it back if requested to. For testers: if you tested 14.24, don't test 15.0, they are exactly the same. Obsidian 15.0 vs 14.0 (LTC) Elo   | 27.21 +- 4.01 (95%) Conf  | 40.0+0.40s Threads=1 Hash=128MB Games | N: 7012 W: 2009 L: 1461 D: 3542 Penta | [7, 576, 1796, 1116, 11] Obsidian 15 SE vs - other engines: Integral 7.0.0 26/52 +0 52 Games Stockfish 17 16.5/34 -1 34 Games Horsie 1.0.6 JA 10.5/20 +1 20 Games YuliGM 19 10/20 +0 20 Games Yuliana 6.0 9/18 +0 18 Games SugaR AI 2.70 8/16 +0 16 Games Cool Iris 14 8/16 +0 16 Games Artemis 7TR 8/16 +0 16 Games HypnoS ++ 1.02 8/16 +0 16 Games Raid 3.7 TR 7.5/15 +0 15 Games SF-PRO2 05.02.2025 6.5/14 -1 14 Games Viridithas 16.0.0 7.5/12 +3 12 Games JudaS ++ 1.02 6/12 +0 12 Games Yuliana 7.0 6/12 +0 ...

Aku 2025.30.5 JA - new version chess engine (Windows and Linux)

Image
Aku Chess Engine - UCI chess engine Author: Hoavu-cs v.2025.05.30: Adjust null move pruning threshold and reduction (from 4 to 3 and from 3 to 3 + depth / 4) -------------------------------------------------- Results of NewAku vs OldAku (30+0.3, 1t, 64MB, imbalanced_book.epd): Elo: 18.71 +/- 9.33, nElo: 28.38 +/- 14.13 LOS: 100.00 %, DrawRatio: 42.08 %, PairsRatio: 1.37 Games: 2324, Wins: 594, Losses: 469, Draws: 1261, Points: 1224.5 (52.69 %) Ptnml(0-2): [48, 236, 489, 321, 68], WL/DD Ratio: 0.39 LLR: 2.94 (100.0%) (-2.94, 2.94) [-5.00, 0.00] -------------------------------------------------- Aku 2015.22.5 JA chess engine - position test The engine was tested in the difficult position you see in the diagram. Here 1.Rg4+ wins. However, Aku, after 10 minutes of consideration, still proposed the incorrect 1.Bd5. How strong is the new version of the Aku 2015.22.5 JA chess engine? The new version of the Aku 2015.22.5 JA chess engine has been released recently and has not been tested ye...

Reckless 0.8.0 dev-2fe JA wins Arena Challenger Engines Tournament, 20-29.05.2025

Image
Reckless 0.8.0 dev-2fe JA won the tournament of new versions of chess engines aspiring to the chess elite. In second place Turbulence 0.0.7 and in third Bagatur 5.2 JA . More results soon. CEDR  646.387 games download  (01.05.2025) 💾  Tournament download with 380 games 💻Asus Zenbook, AMD Ryzen 9 6900HX, RAM 32,0 GB 🖬 GUI-Arena 👍 @chessenginesdiary Country - Poland, City -  Malbork   🕓 Time 3'+3"  P Engine CEDR Score (Tie) Chg +/-/= 1: Reckless 0.8.0 dev-2fe JA 3360 35.5 / 38 610.75 +19 (+33 -0 =5) 2: Turbulence 0.0.7 2400 32.5 / 38 519.75 +190 (+29 -2 =7) 3: Bagatur 5.2 JA 3118 32.0 / 38 516.50 +0 (+27 -1 =10) 4: Illumina 3 2400 30.0 / 38 463.25 +163 (+25 -3 =10) 5: Leorik 3.1.3 JA 2998 30.0 / 38 449.25 -4 (+26 -4 =8) 6:...

Artemis 15TR wins Strong Engines Tournament (Tests by Jörn Gronemann, 2025.05.29)

Image
The big tournament of the strongest chess engines was won by Artemis 15TR . Second place was taken by Stockfish 17.1 and third by Killfish PB 200525 . Obsidian dev-16.01 achieved very good results, it is not far from the best. More results soon. GPU NVIDIA  RTX 3060 ti. GUI-Banksia, For Every Engine only 1 CPU and 512 Hashtables, Book - 4 GM mvs book,  6TBs. 🕒🕒Time 3'+3"     Tournament download with 1360 games   💾 All games CEDR  646.387 games download  (01.05.2025)   @chessenginesdiary P Engine Score (Tie) +/-/= 1: Artemis 15TR 82.0 / 160 6547.25 (+4 -0 =156) 2: Stockfish 17.1 81.5 / 160 6510.50 (+6 -3 =151) 3: Killfish PB 200525 81.5 / 160 6506.25 (+6 -3 =151) 4: DarkSeid v6.0 81.5 / 160 6504.75 (+5 -2 =153) 5: CorChess 20250...

Omega 6-M - new chess engine by Solista (Stockfish derivatives)

Image
Omega - new chess engine by Solista (Stockfish derivatives) Author:  This version includes the Random Op. MultiPV mode.  Two variants are preset, adjustable from 1-500. This examines multiple variants in parallel and merges them into a single variant after reaching a certain depth.  This makes the engine tactically more powerful, but slower overall in its search. Therefore, I've made the search somewhat faster than in the standard version. Those with fast computers can play Blitz well with the default settings. Those with slow computers should reduce the number of variants in the UCI options to 1.  Then the engine plays largely like the standard version and is even slightly faster in its search. File downloads: avx2, bmi2, sse41. Omega 6-M download engine from page: https://solistachess.jimdosite.com/solista-news/

Cool Iris v19 wins Elite Engines Tournament, 26-28.05.2025

Image
Another tournament of elite chess engines conducted in the Fritz 20 program. Cool Iris v19 won and the Artemis 15TR engine in second place. Places 3-5 with the same result were shared by the SF PB 250525, SF-PRO2 02.05.2025 and Stockfish dev-20250523 engines. More results soon on our website. CEDR  646.387 games download  (01.05.2025) 💾  Tournament download with 182 games 💻Asus Zenbook, AMD Ryzen 9 6900HX, RAM 32,0 GB 🖬 GUI-Fritz 20 👍 @chessenginesdiary Country - Poland, City -  Malbork   🕓 Time 3'+3"  P Engine Score (Tie) +/-/= 1: Cool Iris v19 15.0 / 26 185.25 (+4 -0 =22) 2: Artemis 15TR 14.5 / 26 179.00 (+3 -0 =23) 3: SF PB 250525 14.0 / 26 176.00 (+2 -0 =24) 4: SF-PRO2 02.05.2025 14.0 / 26 176.00 (+2 -0 =24) 5: Stockfish dev-20250523 14.0 / 26 176.00 (+2 -0 =...

Cinder 0.2.0 - new version chess engine

Image
  Cinder -  UCI chess engine  Author: Bruno Dutra v.0.2.0: SPRT -------------------------------------------------- Results of dev vs base (1+0.01, 1t, 16MB, UHO_Lichess_4852_v1.epd): Elo: 130.39 +/- 4.49, nElo: 217.01 +/- 6.81 LOS: 100.00 %, DrawRatio: 28.72 %, PairsRatio: 9.04 Games: 10000, Wins: 4719, Losses: 1133, Draws: 4148, Points: 6793.0 (67.93 %) Ptnml(0-2): [18, 337, 1436, 2459, 750], WL/DD Ratio: 1.12 -------------------------------------------------- What's Changed Implement time management based on node count per pv by @brunocodutra in #743 implement time management based on score stability by @brunocodutra in #744 Implement the SPSA algorithm to auto-tune hyper-parameters by @brunocodutra in #763 Add support for Syzygy tablebases by @brunocodutra in #777 How powerful is the Cinder chess engine? The previous version of this engine in the Chess Engines Diary - CEDR ranking achieved a rating of 3580 . This gives it 41st place on the list from May 1, 2025. This v...

Tarnished 2.1 JA - new version chess engine

Image
Tarnished - UCI chess engine Author: Anik Patel (Bobingstern) What has changed in the new version Tarnished 2.1? This version comes with bugfixes, new search features, a bigger and stronger NNUE, and lots of elo! Elo   | 281.87 +- 18.77 (95%) Conf  | 40.0+0.40s Threads=1 Hash=64MB Games | N: 1010 W: 713 L: 36 D: 261 Penta | [0, 1, 49, 232, 223] https://chess.n9x.co/test/1680/ NNUE Tarnished 2.1 now makes use of a (768->512)x2->1x8 architecture, trained on ~650m positions (~1b unfiltered) An inferencing bug causing memory errors has now been fixed thanks to sanitizers Search Improvements SEE Move ordering Fixed improving heuristic QS Check Evasions Singular Extensions Double Extensions Negative Extensions Continuation History Capture History Static evaluation Correction History (Pawn Hash) Fixed mate scores Communication UCI info score is not yet normalized but the option UCI_ShowWDL may be set to true to display WDL values making use of Stockfish's normalization pa...